Loading...
Publication
Integração da inteligência artificial no diagnóstico precoce de lesões potencialmente malignas na cavidade oral: scoping review
Authors
Advisor(s)
Abstract(s)
Introdução: O cancro oral (CO) é uma patologia com um impacto bastante negativo devido ao mau prognóstico quando diagnosticado tardiamente. As lesões orais potencialmente malignas (LOPM) apresentam um risco aumentado de evoluir para cancro, tornando-se crucial a sua deteção precoce. No entanto, os métodos de rastreio convencionais são ainda limitados. A inteligência artificial (IA) está a emergir como uma tecnologia promissora para melhorar o diagnóstico precoce deste tipo de lesões, podendo contribuir de uma forma positiva para a sobrevida dos doentes. Objetivo: Esta revisão explora a utilização da IA no diagnóstico da LOPM, avalia as suas vantagens em relação aos métodos convencionais e identifica os desafios e as limitações da sua integração clínica. Metodologia: Foi realizada uma scoping review seguindo a metodologia PRISMA, incluindo artigos publicados entre janeiro de 2000 e março de 2024 na PubMed, Web of Science e na Biblioteca Virtual em Saúde. Os termos de pesquisa incluíram IA, diagnóstico precoce e lesões orais, cancro oral, diagnóstico precoce. Resultados: Os algoritmos de IA, especialmente as redes neuronais convolucionais, apresentam uma elevada precisão e especificidade na deteção de LOPM e CO. Os modelos de IA demonstraram uma sensibilidade de até 95%, superando frequentemente os métodos convencionais. As aplicações móveis baseadas em IA, como o MobileNet e o VGG19, são eficazes no diagnóstico de lesões orais em ambientes com poucos recursos. Os principais desafios incluem a necessidade de conjuntos de dados alargados e diversificados, a resistência à adoção de tecnologias de IA e questões éticas e regulamentares. Conclusões: A IA mostra um potencial significativo para melhorar o diagnóstico precoce da LOPM, com benefícios notáveis em termos de precisão e rapidez. No entanto, são necessários esforços para superar os desafios relacionados com a formação clínica, a qualidade dos dados e as questões éticas.
Introduction: Oral cancer (OC) is a pathology with a very negative impact due to the poor prognosis when diagnosed late. Oral potentially malignant lesions (OPML) have an increased risk of developing into cancer, making early detection crucial. However, conventional screening methods are still limited. Artificial intelligence (AI) is emerging as a promising technology to improve the early diagnosis of this type of injury, potentially contributing positively to patient survival. Objective: This review explores the use of AI in diagnosing OPML, evaluates its advantages over conventional methods, and identifies challenges and limitations of its clinical integration. Methodology: A scoping review was carried out following the PRISMA methodology, including articles published between January 2000 and March 2024 in PubMed, Web of Science and the Virtual Health Library. Search terms included AI, early diagnosis and oral lesions, oral cancer, early diagnosis. Results: AI algorithms, especially convolutional neural networks, show high accuracy and specificity in detecting LOPM and CO. AI models have demonstrated sensitivity of up to 95%, often outperforming conventional methods. AI-based mobile applications such as MobileNet and VGG19 are effective in diagnosing oral lesions in low-resource settings. Key challenges include the need for large and diverse data sets, resistance to the adoption of AI technologies, and ethical and regulatory issues. Conclusions: AI shows significant potential to improve early diagnosis of LOPM, with notable benefits in terms of accuracy and speed. However, efforts are needed to overcome challenges related to clinical training, data quality and ethical issues.
Introduction: Oral cancer (OC) is a pathology with a very negative impact due to the poor prognosis when diagnosed late. Oral potentially malignant lesions (OPML) have an increased risk of developing into cancer, making early detection crucial. However, conventional screening methods are still limited. Artificial intelligence (AI) is emerging as a promising technology to improve the early diagnosis of this type of injury, potentially contributing positively to patient survival. Objective: This review explores the use of AI in diagnosing OPML, evaluates its advantages over conventional methods, and identifies challenges and limitations of its clinical integration. Methodology: A scoping review was carried out following the PRISMA methodology, including articles published between January 2000 and March 2024 in PubMed, Web of Science and the Virtual Health Library. Search terms included AI, early diagnosis and oral lesions, oral cancer, early diagnosis. Results: AI algorithms, especially convolutional neural networks, show high accuracy and specificity in detecting LOPM and CO. AI models have demonstrated sensitivity of up to 95%, often outperforming conventional methods. AI-based mobile applications such as MobileNet and VGG19 are effective in diagnosing oral lesions in low-resource settings. Key challenges include the need for large and diverse data sets, resistance to the adoption of AI technologies, and ethical and regulatory issues. Conclusions: AI shows significant potential to improve early diagnosis of LOPM, with notable benefits in terms of accuracy and speed. However, efforts are needed to overcome challenges related to clinical training, data quality and ethical issues.
Description
Keywords
Inteligência artificial Diagnóstico precoce Lesões potencialmente malignas Cancro oral Artificial intelligence Early diagnosis Potentially malignant lesions Mouth cancer